Retrieval of images, especially for example facial images, from a relatively large collection of reference images remains a significant problem. It is generally considered impractical for a user to simply browse a relatively large collection of images, for example thumbnail images, so as to select a desired image. Traditionally, images have been indexed by keyword(s) allowing a user to search the images based on associated keywords, with the results being presented using some form of keyword based relevancy test. Such an approach is fraught with difficulties since keyword selection and allocation generally requires human tagging, which is a time intensive process, and many images can be described by multiple or different keywords.
Traditional approaches to content based image retrieval have borrowed techniques from medical and industrial imaging, where content based techniques are used to identify defects, tumours or other characteristics within an image. However, these techniques are not cognisant of the photographic composition inherent in many images, for example images found in the press and media.
There is a need for a method, system, computer program product, article and/or computer readable medium of instructions which addresses or at least ameliorates one or more problems inherent in the prior art.
The reference in this specification to any prior publication (or information derived from the prior publication), or to any matter which is known, is not, and should not be taken as an acknowledgment or admission or any form of suggestion that the prior publication (or information derived from the prior publication) or known matter forms part of the common general knowledge in the field of endeavour to which this specification relates.